f



Descriptive statistics for sample included in multiple regression analysis

I'm running a hierarchical multiple regression analysis and from a dataset =
containing roughly 5,000 participants. When I run the analysis, due to miss=
ing data for some participants, the analysis is done using about 4,000 part=
icipants.

My query is about how do I obtain descriptive statistics just for the sampl=
e of participants who were included in the analysis?

Different participants were excluded for different pieces of missing data a=
cross 15 predictor variables (some dummy coded), so I can't go through the =
data to find each participant with a missing piece of data to individually =
exclude them from the descriptive analysis.

Thank you for any help!
0
cosgrag
8/1/2016 12:10:35 PM
comp.soft-sys.stat.spss 5679 articles. 0 followers. Post Follow

2 Replies
215 Views

Similar Articles

[PageSpeed] 4

On 01/08/2016 8:10 AM, cosgrag@tcd.ie wrote:
> I'm running a hierarchical multiple regression analysis and from a dataset containing roughly 5,000 participants. When I run the analysis, due to missing data for some participants, the analysis is done using about 4,000 participants.
>
> My query is about how do I obtain descriptive statistics just for the sample of participants who were included in the analysis?
>
> Different participants were excluded for different pieces of missing data across 15 predictor variables (some dummy coded), so I can't go through the data to find each participant with a missing piece of data to individually exclude them from the descriptive analysis.
>
> Thank you for any help!
>

Something like the following ought to do the trick.

* Replace Y and X1...XP below with variables in your model.
COMPUTE CompleteCase = NMISS(Y,X1,X2,...Xp) = 0.
FORMATS CompleteCase (F1).
FREQUENCIES CompleteCase. /* Frequency for 1 should = your N.

USE ALL.
FILTER BY CompleteCase.
* Syntax to get descriptive stats here.
USE ALL.
FILTER OFF.

-- 
Bruce Weaver
bweaver@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/Home
"When all else fails, RTFM."
0
Bruce
8/1/2016 2:48:58 PM
On Mon, 1 Aug 2016 10:48:58 -0400, Bruce Weaver <bweaver@lakeheadu.ca>
wrote:

>On 01/08/2016 8:10 AM, cosgrag@tcd.ie wrote:
>> I'm running a hierarchical multiple regression analysis and from a dataset containing roughly 5,000 participants. When I run the analysis, due to missing data for some participants, the analysis is done using about 4,000 participants.
>>
>> My query is about how do I obtain descriptive statistics just for the sample of participants who were included in the analysis?
>>
>> Different participants were excluded for different pieces of missing data across 15 predictor variables (some dummy coded), so I can't go through the data to find each participant with a missing piece of data to individually exclude them from the descriptive analysis.
>>
>> Thank you for any help!
>>
>
>Something like the following ought to do the trick.
>
>* Replace Y and X1...XP below with variables in your model.
>COMPUTE CompleteCase = NMISS(Y,X1,X2,...Xp) = 0.
>FORMATS CompleteCase (F1).
>FREQUENCIES CompleteCase. /* Frequency for 1 should = your N.
>
>USE ALL.
>FILTER BY CompleteCase.
>* Syntax to get descriptive stats here.
>USE ALL.
>FILTER OFF.

[Does USE ALL do anything that is not done by "FILTER OFF"?]

Assuming that your regression seems to have some value, you 
should also look at the characteristics of the excluded cases; 
and (perhaps) contrast the two groups. 

Do the values for excluded cases support the assumption that the 
Missing is at-random across cases?  If not, you have a problem for
drawing inferences.  That is why there is a large literature on 
dealing with Missing in various ways -- using extra dummy variables; 
schemes to replace missing values; the consequences of dropping 
variables or cases; whatever. 

-- 
Rich Ulrich 

0
Rich
8/1/2016 4:42:14 PM
Reply: